[None][perf] Allow CUDA-core NVFP4 GEMM for M up to 16#16345
Open
mihai-chiorean wants to merge 4 commits into
Open
[None][perf] Allow CUDA-core NVFP4 GEMM for M up to 16#16345mihai-chiorean wants to merge 4 commits into
mihai-chiorean wants to merge 4 commits into
Conversation
Signed-off-by: Mihai Chiorean <mihai.v.chiorean@gmail.com>
Contributor
📝 WalkthroughWalkthroughCUDA Core NVFP4 GEMM eligibility is expanded from M ≤ 8 to M ≤ 16. Runtime documentation, the shape limit, and FP4 linear tests are updated to cover the larger dimension. ChangesNVFP4 CUDA Core eligibility
Estimated code review effort: 2 (Simple) | ~10 minutes Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches🧪 Generate unit tests (beta)
Comment |
yizhang-nv
previously approved these changes
Jul 15, 2026
Contributor
Author
|
@yizhang-nv what happened? why dismiss? any feedback? |
mikeiovine
approved these changes
Jul 23, 2026
Contributor
Author
|
@mikeiovine thanks for the stamp! any chance you could trigger CI too? |
asfiyab-nvidia
approved these changes
Jul 23, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
M <= 8toM <= 16, matching the C++cudaCoreGemmTemplateMaxMlimit.cuda_core-only configurations still fail fast when a shape exceeds CUDA-core support.Performance
Spark SM121, Qwen3.6 35B A3B NVFP4, no-MTP graph+overlap smoke (with my other patches to get Qwen3.6 35B A3B NVFP4 to run):
Testing
uvx pre-commit run --files tensorrt_llm/_torch/custom_ops/torch_custom_ops.py tests/unittest/_torch/thop/parallel/test_fp4_linear.pypytest -q tests/unittest/_torch/thop/parallel/test_fp4_linear.py -k test_fp4_linear_cuda_corevia TensorRT bindings bootstrap: 4 passedpytest -q tests/unittest/_torch/thop/parallel/test_fp4_linear.py -k "test_fp4_gemm_bias_per_backend and cuda_core"via TensorRT bindings bootstrap: 1 passed, 3 skippedSummary by CodeRabbit
New Features
Mup to 16.Bug Fixes
M=16workloads and bias parity checks.